Objective To explore the statistical analysis method of small-sample micronuclei data by simulation studies.Mehods To simulate small-sample micronuclei data, then compare the size and power of the test methods in literature for the micronuclei assay analysis.Results Compare the typeâ… error and statistical power of several different small-sample anaysis approaches for micronuclei data. statistical analysis methods ,such as Poisson exact test ,Fisher exact test , Negative Binomial Exact Test, t test ,t test after transformation, rank sum exact test, are considered. The power of tests increase withδand n.δincrease with cell and positive rate.statistical power of Poisson exact test is the highest .Fisher exact test and Negative Binomial Exact Test is lower than Poisson exact test,and higher than other statistical analysis methods。But under negative binomial distribution, typeâ… error are large in Poisson exact test and Fisher exact test, while it is small in Negative Binomial Exact Test. Under Poisson distribution, typeâ… error are large in Poisson exact test, Negative Binomial Exact Test and Fisher exact test are similar.Both of them are smaller than level of the test or closed to it.Other statistical analysis methods increase with n.Conclusion the power of the Negative Binomial Exact is higher in spite of the lower typeâ… error...
|